Fast and robust human detection method in range map of complex environment

Abstract

With increasing need of human-computer interfaces, pose estimation or gesture recognition has drawn a lot of attentions. Since humans are one of the subjects of human-computer interfaces, human detection is required as a preprocessing step for human-computer interaction system. In this paper, we propose a fast and robust human detection method by using a range map. Instead of finding exact shape of human in the range map, the proposed method uses static background modeling and labeling technique. The proposed method is robust against self-occlusions and can separate undesirably merged labels. The experiments show that the proposed method efficiently produces foreground regions within 10 ms. Due to its low complexity, the proposed method can be extended to restricted environments such as mobile or embedded systems.